Bill > A08833

This bill introduces the "Understanding Artificial Intelligence Act" in New York, which establishes strict liability for developers of large-scale artificial intelligence (AI) models when their systems cause injuries to non-users. Specifically, the bill defines a "covered model" as an AI system trained using computing power exceeding a certain threshold and costing over $100 million, and a "developer" as the entity performing the initial model training. Under this legislation, developers would be legally responsible for injuries caused by their AI systems if the system's conduct would constitute negligence, an intentional tort, or a crime if performed by a human, and if such conduct was not intended or reasonably anticipated by the user or any intermediary who modified the model. The bill also introduces a rebuttable presumption about the AI system's mental state by comparing its actions to those a human might take under similar circumstances. Developers can potentially avoid liability by proving they met the standard of care applicable to humans performing the same function or by demonstrating that the system's failure was a "capabilities failure" that would not meet the legal standards for negligence or intentional harm. This proposed law represents a significant step in establishing legal accountability for AI system developers and addressing potential harmful consequences of advanced AI technologies.

Business and Industry

https://www.nysenate.gov/legislation/bills/2025/A8833